The myth of “joins don’t scale”
So I was reading this….
Some of the responses are just not educated (similar to people saying 640k of memory is all you need).
You don’t need to denormalized (that’s a bad 1990s concept when we didn’t have the technology).
What is needed is new technology (SIMD), more JOIN algorithms, features like bucketing and partitioning, etc etc. Anyhow it doesn’t really matter since you can have a lot of feature stuffing into a product, what matters is real world usage like how AirBnB use StarRocks and they can do 3 table (billions, billion millions of rows) with 5 JOINS that used to take minutes and now they can do it in ~4 seconds. Read more about it at https://www.starrocks.io/blog/airbnb-builds-a-new-generation-of-fast-analytics-experience-with-starrocks
